IEEE Transactions on automatic control vol:45 issue:2 pages:364-369
This paper presents a frequency domain solution to the system identification problem with missing data in the input and output signals. No particular pattern for the missing data is assumed. The approach does not require a (parametric) model for the input signal, works for any model structure (ARX, ARMAX, OE, errors-in variables, etc.), and can be applied to discrete time as well as continuous time models. The key idea is to treat the missing data as unknown parameters.